Zeus computing board report
Download
1 / 8

ZEUS Computing Board Report - PowerPoint PPT Presentation


  • 77 Views
  • Uploaded on

ZEUS Computing Board Report. Zeus Executive Meeting Wesley H. Smith, University of Wisconsin on behalf of the ZEUS Computing Board: This talk is available on: http://www.hep.wisc.edu/wsmith/zeus/smith_compute_feb06.pdf. Membership: The offline coordination team

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' ZEUS Computing Board Report' - tansy


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Zeus computing board report
ZEUS Computing Board Report

Zeus Executive Meeting

Wesley H. Smith, University of Wisconsin

on behalf of the ZEUS Computing Board:

This talk is available on:

http://www.hep.wisc.edu/wsmith/zeus/smith_compute_feb06.pdf

  • Membership:

    • The offline coordination team

      • Analysis software & MC: A Geiser

      • Reconstruction, code mgmt & data base: U. Stoesslein

      • User support and infrastructure: K. Wrona

    • Five external members

      • J. Ferrando, T. Haas, M. Kuze, R. Mankel, W. Smith (Ch.)

  • First Meeting: February 27, 2006


Zcb mandate
ZCB Mandate

  • ZCB will normally meet 3 times a year to coincide with the ZEUS collaboration meetings.

  • ZCB is charged with the following:

    • Review the current status and short term planning of ZEUS offline. Recommend changes and new initiatives where necessary.

    • Review the progress of the medium term planning as outlined in ZEUS-note 03-024 (ZEUS Computing Strategy: an update for years 2004-2007). Recommend changes and new initiatives where necessary.

    • Towards the end of 2006, formulate a plan for computing beyond 2007.

  • ZCB Chair reports to ZEUS executive session at the ZEUS collaboration meetings.


Zcb planning
ZCB Planning

  • Decide that the first step is to establish the context by which the present activities and future plans can be evaluated.

  • Take a view of the long range plans and establish what steps need to be taken now to prepare ourselves for the rest of ZEUS physics.

  • Review the salient issues to understand what we need to look at in our next meetings in order to make recommendations to the ZEUS executive:

    • Present status of computing systems & resources

    • Data & MC processing

    • Long term physics analysis & file storage


Computing hardware evaluation info from k wrona
Computing Hardware Evaluation(info. from K. Wrona)

  • Present Computing is in good shape

    • Mass storage costs are manageable

      • Maintenance (silos, drives, servers by DESY-IT -- Zeus share) cost of 70K€

      • Cartridge cost of 60K€ for 850 cartridges for new Zeus data, assumes:

        • Long term storage of only 1 MDST version

        • restricting amount of MC (see later)

        • reprocessed data stored on recovered tapes (old MDSTs)

    • CPU capacity is adequate for analysis & reconstruction

      • ZARAH farm: 184 CPUs in 92 dual CPU nodes

        • Including CPU upgrades planned this year to replace oldest nodes

    • Central File server has enough capacity

      • Remove objectivity, use zeslite

      • Maintenance is 6.6K€/yr

    • Workgroup servers are presently sufficient

      • OK if institutes continue to provide disk space and servers for their members

      • Central WG servers being reduced in number, but higher power


Evaluate software cache grid info from k wrona
Evaluate: Software, Cache, GRID(info. from K. Wrona)

  • Present SW evolution needs to continue

    • PC Farms & WG servers migrate DL5  SL3

      • Widespread use & support thru 2008

    • Analysis SW migrate PAW  ROOT5 (ZEUS default)

      • Good for training of students, postdocs & compatible with latest tools

      • Keep “legacy” PAW capability

  • We need to invest in dCache

    • Replace old dCache servers & add more

      • Add 36 TB using better, more cost effective hardware supported by DESY-IT.

  • Increase use of GRID resources for MC & analysis

    • Plenty of GRID resources for ZEUS MC generation

    • If shortage of analysis computing, move more analysis to the GRID

      • First prototype analysis working on the GRID -- some analyses are appropriate

      • Not appropriate for reconstruction, reprocessing, etc.

      • Need to have a computing reserve beyond ZEUS resources alone for analysis

    • Need to maintain/enhance ZEUS operation on GRID - dynamic target!

    • May need “deals” with ZEUS groups w/GRID resources


Computing finances in the future
Computing Finances in the Future

  • 2006 ZARAH budget reduced 160K€ 140K€by VAT

  • Assumed 2007 budget is the same

  • Have no clue about 2008 budget and beyond

    • May be significantly reduced?

    • Need to start the discussion now…critical to planning

  • In what follows, assume as little additional funding as possible…basically just support for what we have with no enhancements.


Data processing info from u stoesslein
Data Processing(info. From U. Stoesslein)

  • Estimate ultimate ZEUS data size

    • Start with 04/05 e- data

      • TLT  = 1 b, 41% efficiency for 346 HERA running days @ 1 pb-1/day

      • RAW 15 TB & MDST 7 TB

    • Estimate 06 data

      • TLT  = 0.75 b, 50% efficiency for 310 HERA running days @ 1 pb-1/day

      • RAW 13 TB & MDST 7 TB

    • Estimate 07 data

      • TLT  = 0.75 b, 60% efficiency for 180 HERA running days @ 1 pb-1/day

      • RAW 10 TB & MDST 4.5 TB

  • Reprocessing

    • Need resources if plan to start reprocessing data as soon as possible

      • Suppose start reprocessing of 06 data in summer -- keep 6 months interval max

      • Allows completion of all reprocessing by end of 2007

      • Implies reprocessing of 04/05 data also starting in summer 2006

      • Implies greater demand on computing since reconstruction/reprocess/analyses in parallel

        • Seek additional resources for analysis from GRID

    • What about “grand reprocessing”?

      • If we follow previous ZEUS timeline -- ready in middle 09 --- realistic?


Analysis mc info from a geiser
Analysis & MC(info from A. Geiser)

  • Analysis Plan Assumptions

    • MC = 5 x Data

      • MC event records are longer than data records (30%)

    • Move to common ntuples

      • Based on ORANGE routines “certified & stable”

      • Make analyses more efficient (common tools, systematics, etc.)

      • Not realistic to have each analysis keep own MC or own ntuples

  • MC Sample Size

    • Based on available storage, limit 2006 sample to ~ 75 TB

      • Corresponds to ~ 625 M events at 120 kb/event

      • Need to generate at ~ 12.5 M events/week -- OK on GRID

  • Use of common ntuples to save resources

    • Physics groups maintain “in common”

      • Move resources from individual ntuple sets to common ntuple sets

      • Some groups (HFL) have done this, others (QCD) not.

      • Eventually, remove need for MDST? (Size of ntuple?)

      • As a test, temporarily allocate 1TB ZEUS central resources for common ntuple


ad